The release of hazardous material in a subway station is a dangerous situation for passengers and staff. To achieve an acceptable level of risk, a correct marking of emergency exits and escape routes is essential., For this purpose, underground air dispersion, which depends on source characteristics, the construction layout of the station, and meteorological conditions, must be known. In this paper a micro-scale wind model is combined with a dispersion model to calculate the spread of pollutants underground as well as aboveground for specific scenarios. In particular, the effect of emission temperature on the contamination of the subway facilities was studied. Completely different spots with high concentrations have been calculated for hot and cold sources. Depending on the scenario, an escape route is not always safe, underlining the need for a situation-related adaptive marking system. Further information on the optimisation of a safe way out is provided by the results of an agent-based simulation model. Depending on the number and on the individual characteristics of the agents, the contamination level can be estimated, which could help in the prioritisation process between alternative escape routes.